Data-driven statistical modeling of a cube regrasp

R. Paolini,M. T. Mason

DOI: https://doi.org/10.1109/IROS.2016.7759397

2016-10-01

Abstract:Regrasping is the process of adjusting the position and orientation of an object in one's hand. The study of robotic regrasping has generally been limited to use of theoretical analytical models and cases with little uncertainty. Analytical models and simulations have so far proven unable to capture the complexity of the real world. Empirical statistical models are more promising, but collecting good data is difficult. In this paper, we collect data from 3300 robot regrasps, and use this data to learn two probability functions: 1) The probability that the object is still in the robot's hand after a regrasp action; and 2) The probability distribution of the object pose after the regrasp given that the object is still grasped. Both of these functions are learned using kernel density estimation with a similarity metric over object pose. We show that our data-driven models achieve comparable accuracy to a geometric model and an off-the-shelf simulator in classification and prediction tasks, while also enabling us to predict probability distributions.